STOCHASTIC MODELING OF SHORT-TERM VARIATI ONS OF SEA-LEVEL IN THE ST-LAWRENCE ESTUARY, CANADA

Hourly sea level records taken at Quebec-Lauzon in the St. Lawrence es tuary, Canada, are analysed both in frequency domain from 1970 to 1979 and time domain during 1973. Periodic variations, identified by spect ral analysis and adjusted by harmonic regression, explain 90 to 95% of the total sea level variability. The residual (non tidal) variations of sea level, stochastic in nature, are responsible of great amplitude s of seiches and storm surges. Such series, representing less than 10% , is tested for its stationnarity and randomness and fitted by AutoReg ressive-MovingAverage (ARMA) model. The long-period variations (2 to 2 8 days) correspond to atmospheric pressure and winds. The short-period variations (2 hours to 1 day) call be attributed to seich-like motion s, semidiurnal and diurnal atmospheric tides and inertial oscillations . River discharge, atmospheric pressure and winds contributed 29%, 8.1 % and 8.9% respectively to the monthly residual sea level variations. The longitudinal component of sea surface wind (U), parallel to the co ast, acts more on residual sea level than the transversal component (V ), Perpendicular to the shore, and contributes about 7%. The relations hip between residual sea level and atmospheric pressure field is estim ated as -1.5 cm.(hPa)(-1) (+/- 0.3 cm.(hPa)(-1)).